Periodic Reporting for period 2 - PharmScreen2 (Holistic Quantum Mechanics Based Platform to Accelerate Drug Discovery)
Periodo di rendicontazione: 2021-09-01 al 2022-08-31
Life expectancy has reached a plateau in most developed countries and the incidence of chronic and life-threatening diseases is increasing, highlighting the need of novel therapies to cure them. However, the productivity of drug discovery R&D has been steadily declining. In fact, the inflation-adjusted research and development costs for a new drug has increased 100-fold between 1950 and 2010—roughly doubling every 9 years. In contrast to the semiconductor industry, where Moore’s law has shown that chip integration / performance doubles every two years, this decline in pharma R&D has been named Eroom’s law (“Moore” in reverse order). Nowadays, the costs of new drug development have increased from €1 billion in 2000 to €2.58 billion in 2015. The increase is due in part to the difficulty to find New Molecular Entities (NME) subject to be patented. This poses a serious threat to continue developing new therapies for unmet medical needs.
Drug development projects are long processes that take approximately 15 years from the moment a molecular target is discovered to the drug being approved to be used by patients. The process is divided in a pre-clinical and clinical phase with subsequent sub-phases, and in each one of those, huge amounts of money are invested. However, the success rates are very low since the suitability of molecules is difficult to predict. The goal of a preclinical drug discovery program is to deliver one or more clinical candidate molecules, each of which has enough evidence of biologic activity as well as sufficient safety and drug-like properties so that it can be entered into human testing. If the results of a phase are not optimal enough, the preclinical sub-phases must be repeated depending on the cause of failure, meaning higher costs, lost time and longer suffering patients.
Late preclinical and clinical stages are very costly and failing at these points significantly increases the project duration and investment. For this reason it is crucial for the pharmaceutical/biotech sector to move towards a lean drug discovery model where more design cycles are performed in the early stages of the development process (agile, flexible and cheap), accurately predicting, testing, learning and failing fast (if necessary) and feeding subsequent expensive stages with molecules that have higher chances to become a drug.
Computer-Aided Drug Design (CADD), also known as in-silico drug design, has become a powerful technique in the initial phases of drug discovery but has focused mainly in specific aspects of the drug alone, being the most optimized property the activity against the target of interest. However, the high attrition rates (96%) of drug discovery projects indicates that they are not identifying strong candidate molecules to be used at later stages. New CADD methodologies will have a key role in these initial stages to enable focused compound testing and optimization. Intelligent in-silico screening of huge virtual libraries (virtual screening) enables fast and affordable design cycles that can generate valuable experimental data and learnings to improve an iterative cascade screening process. PharmScreen2, the platform developed in this project, is based on an integrated HPC platform combined with ML and QM algorithms, allowing researchers to perform advanced simulations and to find the best new candidate molecules with a much higher potential to become new drugs.
Drug development projects are long processes that take approximately 15 years from the moment a molecular target is discovered to the drug being approved to be used by patients. The process is divided in a pre-clinical and clinical phase with subsequent sub-phases, and in each one of those, huge amounts of money are invested. However, the success rates are very low since the suitability of molecules is difficult to predict. The goal of a preclinical drug discovery program is to deliver one or more clinical candidate molecules, each of which has enough evidence of biologic activity as well as sufficient safety and drug-like properties so that it can be entered into human testing. If the results of a phase are not optimal enough, the preclinical sub-phases must be repeated depending on the cause of failure, meaning higher costs, lost time and longer suffering patients.
Late preclinical and clinical stages are very costly and failing at these points significantly increases the project duration and investment. For this reason it is crucial for the pharmaceutical/biotech sector to move towards a lean drug discovery model where more design cycles are performed in the early stages of the development process (agile, flexible and cheap), accurately predicting, testing, learning and failing fast (if necessary) and feeding subsequent expensive stages with molecules that have higher chances to become a drug.
Computer-Aided Drug Design (CADD), also known as in-silico drug design, has become a powerful technique in the initial phases of drug discovery but has focused mainly in specific aspects of the drug alone, being the most optimized property the activity against the target of interest. However, the high attrition rates (96%) of drug discovery projects indicates that they are not identifying strong candidate molecules to be used at later stages. New CADD methodologies will have a key role in these initial stages to enable focused compound testing and optimization. Intelligent in-silico screening of huge virtual libraries (virtual screening) enables fast and affordable design cycles that can generate valuable experimental data and learnings to improve an iterative cascade screening process. PharmScreen2, the platform developed in this project, is based on an integrated HPC platform combined with ML and QM algorithms, allowing researchers to perform advanced simulations and to find the best new candidate molecules with a much higher potential to become new drugs.
During the first year of the PharmScreen2 project, Pharmacelera's platform has been improved in multiple ways, all of them aligned with the main goals of the project which are be able to perform accurate searches of a huge and richer chemical space, having an holistic view of molecular properties when searching for new candidate molecules and having a user friendly and easy to integrate interface.
Accurate searches of a huge and richer chemical space
The molecular search algorithms of PharmScreen have been improved to take into account information about the binding site in conjunction with the information of active ligands as a way of improving the accuracy of our 3D field-based approach.
A new methodology has been implemented that enables PharmScreen to perform molecular searches in a molecular space that is orders of magnitude bigger than what has been traditionally explored with CADD tools.
Holistic screening
Pharmacelera's hydrophobicity descriptors have been used in conjunction with machine learning methods to enable the prediction of ADME-Tox properties and Off-target interactions and enable providing a broader view of molecule properties in early drug discovery stages.
Usability and Integration
A new API for PharmScreen has been created to enable a seamless integration with other workflows. This API allows to easily use PharmScreen with Python using a remote workstation or cluster in Amazon Web Services in conjunction with a newly created Python package.
Multiple improvements have been added to the SaaS version of PharmScreen such as an improved graphical user interface with a better molecular viewer to interact with the molecules, a molecule ranking table to visualize the results, and the addition of different types of users.
Accurate searches of a huge and richer chemical space
The molecular search algorithms of PharmScreen have been improved to take into account information about the binding site in conjunction with the information of active ligands as a way of improving the accuracy of our 3D field-based approach.
A new methodology has been implemented that enables PharmScreen to perform molecular searches in a molecular space that is orders of magnitude bigger than what has been traditionally explored with CADD tools.
Holistic screening
Pharmacelera's hydrophobicity descriptors have been used in conjunction with machine learning methods to enable the prediction of ADME-Tox properties and Off-target interactions and enable providing a broader view of molecule properties in early drug discovery stages.
Usability and Integration
A new API for PharmScreen has been created to enable a seamless integration with other workflows. This API allows to easily use PharmScreen with Python using a remote workstation or cluster in Amazon Web Services in conjunction with a newly created Python package.
Multiple improvements have been added to the SaaS version of PharmScreen such as an improved graphical user interface with a better molecular viewer to interact with the molecules, a molecule ranking table to visualize the results, and the addition of different types of users.
PharmScreen2 is a platform that will enable finding new drugs faster and at a lower cost. This will have a significant impact in the way drugs are designed and advancing molecules with lower chances of failure in preclinical and clinical stages.
Our current technology is already capable of exploring a chemical space of a size that was not feasible in the past and adding relevant information to the proposed compounds to minimize potential problems at later stages of the drug development proceess.
Our partnerships with pharmaceutical companies will enable us to validate the technology during the second phase of the project usign high quality experimental data and refine our tool for better performance and usability.
Nowadays is more clear than ever that our society needs to be prepared for unforeseen health threats. This fact, in conjunction with an ageing population that increases the preasure to our healthcare system, shows the need of having a strong R&D ecosystem in the drug discovery field to provide solutions.
We expect that PharmScreen2 will enable researchers tackle these new challenges with more chances of success.
Our current technology is already capable of exploring a chemical space of a size that was not feasible in the past and adding relevant information to the proposed compounds to minimize potential problems at later stages of the drug development proceess.
Our partnerships with pharmaceutical companies will enable us to validate the technology during the second phase of the project usign high quality experimental data and refine our tool for better performance and usability.
Nowadays is more clear than ever that our society needs to be prepared for unforeseen health threats. This fact, in conjunction with an ageing population that increases the preasure to our healthcare system, shows the need of having a strong R&D ecosystem in the drug discovery field to provide solutions.
We expect that PharmScreen2 will enable researchers tackle these new challenges with more chances of success.